Keel Infrastructure Shuts Down U.S. Bitcoin Mining in Pivot to AI Data Centers

Keel Infrastructure has shut down all U.S. Bitcoin mining operations, creating a sharper reversal than the Google News headline suggests. The former Bitfarms now wants investors to value it as an AI and high-performance computing infrastructure developer. Its remaining challenge is substantial. Keel must convert controlled power and proposed campuses into contracted, operating data centers.

The pivot reached a visible milestone on August 10, 2026. Keel reported that its three priority sites were approaching full permitting, with multiple prospective tenants negotiating at each location. It also reported $819 million in liquidity, including cash and unencumbered Bitcoin.

Those figures strengthen Keel’s negotiating position, but they do not finish the transformation. Revenue from continuing legacy operations fell by half from the previous year. No announced long-term AI lease yet supports the company’s new identity.

That gap defines the story. Keel owns infrastructure inherited from cryptocurrency mining, but AI customers demand a different standard of reliability, design, financing, and service. Rivals including Core Scientific, TeraWulf, Applied Digital, and IREN are pursuing similar conversions or expansion strategies.

Keel’s Bitcoin Exit Has Become Physical

Keel is no longer describing an eventual pivot. It has started removing the operating assets that tied its U.S. business to Bitcoin.

The company said it completed the decommissioning of every U.S. Bitcoin mining operation during the second quarter. It described the work as preparation for high-performance computing, or HPC, site construction. HPC refers to tightly connected computing systems built for demanding workloads, including AI training and inference.

This operational change followed a corporate transformation completed on April 1. Bitfarms redomiciled from Canada to the United States and adopted the Keel Infrastructure name. Its shares later began trading under the KEEL symbol on Nasdaq and the Toronto Stock Exchange.

The name change mattered because management attached a different business model to it. Keel would develop power-secured sites and data center facilities for customers, rather than continually investing in Bitcoin mining capacity.

Keel’s rebranding announcement said the shift had started about one year earlier. Chief Executive Ben Gagnon called Keel a pure infrastructure developer and owner focused on sites that support AI compute.

The latest quarter supplied evidence that the physical transition has continued. Keel accepted the first Vertiv modules at Moses Lake, Washington. It also received other long-lead equipment for Moses Lake and Sharon, Pennsylvania.

Long-lead equipment includes components that require extended manufacturing, delivery, or installation schedules. Ordering these systems early can protect a construction timeline. It also commits capital before a tenant contract necessarily guarantees future revenue.

Keel said it was finalizing fiber contracts across Panther Creek, Sharon, and Moses Lake. Fiber provides the high-capacity network connections needed to move data between servers, storage systems, customers, and external networks.

The company also reported progress on zoning, land development approvals, and environmental permits. Panther Creek received conditional land development approval, while Sharon secured zoning and land development approvals. Environmental applications continued across all three priority locations.

These are meaningful development steps because data center capacity cannot be created by installing processors alone. A viable campus needs land, grid access, substations, cooling, network connectivity, permits, and an acceptable construction schedule.

Keel already understands energy-intensive computing through Bitcoin mining. Mining facilities run specialized machines continuously and manage large electrical loads. However, that experience covers only part of an AI tenant’s requirements.

Bitcoin miners can tolerate equipment designs and operating practices that would not satisfy an enterprise computing customer. AI facilities need denser racks, dependable networking, carefully engineered cooling, physical security, and contractual uptime commitments.

Keel’s transition therefore has two layers. It is retiring mining operations while redesigning selected properties for a more demanding customer category. The first layer is visible. The second will be proven only through signed leases, completed facilities, and operating performance.

The distinction also corrects an easy misconception in Google News coverage. Keel has not simply redirected the same computers from Bitcoin calculations to AI models. Cryptocurrency miners use application-specific integrated circuits, which cannot replace general-purpose AI accelerators.

The reusable assets are primarily power, land, utility relationships, buildings, and operating experience. Those assets can shorten parts of the development process. They do not eliminate the need for extensive construction and new computing infrastructure.

Why Google News Headlines Focus on Power

Keel’s strongest asset is not a fleet of AI chips. It is access to locations where large electrical loads can be developed.

Gagnon summarized management’s thesis during the second-quarter release: “Power is the constraint. Everything else is downstream of it.” That statement identifies both Keel’s opportunity and its central dependency.

Keel reports a development pipeline totaling 2.2 gigawatts. A gigawatt equals 1,000 megawatts, and a megawatt measures the rate of electrical power delivery. The pipeline combines energized, secured, and expansion capacity at different development stages.

Those categories should not be treated as interchangeable. Energized capacity is already receiving utility power and being used. Secured capacity has agreements for future delivery. Expansion capacity remains under application, utility study, or evaluation for on-site generation.

Keel’s investor materials list Panther Creek as its flagship development. The Pennsylvania property has 350 megawatts of gross capacity and 336 acres. Management has also discussed a potential increase beyond 500 megawatts.

Sharon has 110 megawatts of gross capacity across 17 acres. Of that total, 30 megawatts were already operating, while substations for another 80 megawatts remained under development in earlier company materials.

Moses Lake is smaller at 18 megawatts across six acres. Its scale makes it less significant than Panther Creek in raw capacity. However, its more advanced equipment activity gives investors a nearer test of execution.

Sherbrooke, Quebec, adds another route into the AI data center market. Keel announced an agreement involving the conditional transfer and operation of 96 megawatts of existing capacity. It also signed an agreement to purchase land for the project.

The second-quarter results show why power attracts attention. Keel says all three priority sites have multiple prospective tenants in negotiations. Available capacity in 2027 remains uncommitted, giving the company room to negotiate terms.

That positioning addresses a genuine industry bottleneck. Large data centers often wait for grid studies, generation resources, transformers, substations, and transmission upgrades. A suitable connection can take longer to obtain than servers take to manufacture.

The International Energy Agency expects electricity consumption from data centers to rise substantially as AI adoption expands. Its energy outlook connects that growth with higher demand from accelerated computing and supporting infrastructure.

Still, a large pipeline does not equal delivered capacity. Expansion megawatts can remain subject to engineering studies, regulatory approval, generation plans, and capital availability. Even secured capacity needs construction before customers can use it.

The terminology can also obscure differences between sites. A small energized facility and a proposed gigawatt campus do not carry the same schedule, risk, or economic value. Investors need to track each location separately.

Keel’s immediate pressure comes from prospective AI tenants, not from Bitcoin competitors. Those customers can compare sites across developers and demand clear answers about power delivery, cooling, network redundancy, construction schedules, and financing.

The company must also satisfy utilities and local authorities. A large data center can influence transmission planning, water use, noise, land use, tax policy, and community development. Technical readiness does not replace public approval.

This pressure explains why Keel has hired executives with infrastructure and commercial experience. Ganesh Aiyer became president in July to lead commercial and expansion work. His assignment places tenant conversion near the center of the company’s next phase.

Power gives Keel a credible entry point. It also creates fixed obligations before customer revenue begins. Deposits, engineering, equipment, permitting, and site work consume capital while negotiations remain uncertain.

For developers, enterprise buyers, and AI product teams, this issue reaches beyond one stock. Compute availability ultimately shapes service capacity, deployment schedules, and model operating costs. The physical layer can constrain plans written entirely in software.

Teams monitoring that shifting landscape need to preserve filings, project updates, and technical claims across many sources. A searchable AI knowledge base can help connect infrastructure announcements with later delivery evidence.

The AI Data Center Pivot Reverses Keel’s Economics

Keel is replacing an operating but volatile mining business with a potentially steadier model that has not yet produced contracted AI revenue.

Bitcoin mining turns electricity into computational guesses that secure a blockchain. Revenue responds quickly to Bitcoin prices, network difficulty, equipment efficiency, and energy costs. Operators can also shut machines down when economics deteriorate.

An AI data center follows another pattern. A developer secures power and land, builds specialized facilities, and signs customers under longer contracts. Revenue can become more predictable, but construction requires more capital and coordination.

This is the core reversal. Keel is giving up near-term mining output to pursue long-term infrastructure contracts. The destination offers a different revenue profile, but the transition creates an exposed period between the two businesses.

That exposure appeared clearly in Keel’s latest financial results. Second-quarter revenue from continuing legacy operations was $30 million, down 50 percent year over year. Management attributed much of the decline to Bitcoin conditions and the April shutdown at Moses Lake.

The quarter produced a $141 million operating loss. That figure included $84 million of non-cash depreciation, which allocates asset costs over their accounting lives. The company reported an operating profit during the comparable 2025 quarter.

Adjusted EBITDA was negative $24 million, compared with positive $7 million one year earlier. Adjusted EBITDA excludes several accounting items, so it should not replace the company’s results under standard accounting rules.

General and administrative spending also rose to $31 million from $19 million. Keel said targeted hiring of senior specialists drove much of the increase as the company moved into project management.

These numbers do not prove the pivot has failed. They show that Keel is absorbing the cost of changing businesses before its targeted contracts begin producing revenue.

The company’s reported liquidity gives it time. As of August 7, Keel held about $698 million in unrestricted cash and $121 million in unencumbered Bitcoin. Together, those assets produced approximately $819 million of liquidity.

Keel strengthened that position by issuing $458 million in convertible senior notes during the quarter. Convertible notes are debt instruments that can convert into equity under defined conditions. They increase financing capacity while creating interest, refinancing, and potential dilution considerations.

The company also sold 1,085 Bitcoin between April 1 and August 7, generating $75 million in proceeds. It retained 1,861 Bitcoin at the end of that period.

This sell-down is another practical sign of separation from the old strategy. Bitcoin has shifted from Keel’s primary operating output toward a funding resource for infrastructure development.

However, liquidity does not remove project-level financing needs. Keel expects leases to support financing structures based on contracted cash flows. Until those leases arrive, the company carries more development and negotiating risk itself.

This transition differs from simply expanding a profitable existing data center platform. Keel must establish credibility with customers while legacy revenue contracts. Prospective tenants will examine whether the company can meet enterprise delivery standards.

The company also faces established data center operators with longer records. Some competitors already operate facilities for cloud and AI customers. Others have secured large contracts that provide lenders with clearer revenue visibility.

Former miners bring valuable power assets, but many are chasing the same opportunity. Core Scientific, TeraWulf, IREN, Applied Digital, Hut 8, and Cipher Mining have all explored AI or HPC capacity alongside cryptocurrency operations.

Core Scientific provided a prominent precedent by expanding hosting agreements for AI workloads. TeraWulf has also directed capacity toward high-density computing customers. These examples show that mining infrastructure can attract AI demand.

They also raise the standard for Keel. Investors can compare signed capacity, customer quality, construction progress, financing, and revenue timing across the group. A large development pipeline alone will not settle that comparison.

Keel’s strategy therefore rests on a specific conversion mechanism. The company secures power, advances permits, negotiates a lease, and then uses contracted cash flows to finance construction. Each step reduces a different risk.

A signed lease reduces customer uncertainty but does not guarantee on-time delivery. Completed financing funds construction but does not remove permitting or supply risks. An operating facility must still meet performance obligations.

That sequence matters more than a broad “Bitcoin to AI” label. Google News can surface the label quickly. Investors and customers must follow whether Keel advances through every link in the development chain.

The Missing AI Leases Still Define the Risk

Keel has demonstrated that it can stop mining, raise capital, and advance sites. It has not yet demonstrated a completed AI infrastructure business.

Management reported active negotiations at Panther Creek, Sharon, and Moses Lake. It also said multiple prospective tenants were negotiating for each location. Those statements signal demand, but negotiations are not contracted revenue.

Keel previously made lease execution across its priority sites a central 2026 objective. A completed agreement would validate customer interest, establish commercial terms, and support project financing. Missing that objective would weaken the pivot’s timeline.

The verification gap is especially important because the company’s capacity figures describe a development pipeline. Some power is energized, some is contractually secured for future delivery, and some remains an expansion opportunity.

Keel’s own regulatory disclosures identify the relevant uncertainties. Its quarterly filing warns that increased AI and HPC development might not become profitable. It also notes that the strategy diverts resources from Bitcoin operations.

The filing identifies capital intensity, construction delays, financing availability, competition, customer concentration, and regulatory approvals as additional risks. These are not generic disclaimers detached from the business model. They correspond directly with Keel’s planned conversion sequence.

Permitting provides one example. Conditional land approval can move a project forward without removing every requirement. Environmental reviews, construction permits, utility work, and local conditions can still affect delivery.

Supply chains provide another. Keel has accepted some Vertiv modules and ordered long-lead equipment. Yet a complete facility requires coordinated delivery across electrical, cooling, networking, structural, security, and control systems.

Power density creates further complexity. AI racks can concentrate much more electricity and heat than older computing layouts. A site must deliver energy reliably and remove heat without damaging performance.

Keel said it was refining designs to improve power density. That work can make a location more attractive to customers. It can also require revised engineering, equipment, and construction plans.

Customer concentration presents a separate risk. Large AI campuses often depend on a small number of tenants because each customer requires substantial capacity. One delayed commitment can therefore change financing and construction schedules.

Contract quality matters as much as contract size. Investors should examine lease duration, credit support, power obligations, construction milestones, termination rights, and responsibility for equipment.

Competition can influence all of those terms. Hyperscalers may build their own campuses or select established developers. Neocloud providers may lease capacity directly and install accelerators. Other former miners may offer sites with comparable power access.

The broader neocloud market adds another uncertainty. A neocloud provides accelerator-focused computing without reproducing every service offered by a large public cloud. Its demand can depend on GPU availability, financing, utilization, and customer growth.

Keel is positioning below that service layer as an infrastructure owner and developer. This focus avoids direct competition with complete cloud platforms. However, it leaves the company dependent on tenants whose own economics can change.

Rapid processor turnover also complicates long-term planning. New accelerators can require different power densities, cooling designs, and networking. Infrastructure must remain useful across several hardware generations.

Community acceptance cannot be assumed either. Data centers can bring investment and tax revenue, but they also raise concerns about electricity, water, noise, and land use. Local resistance can add conditions or delay approvals.

Keel’s North American concentration reduces some geographic complexity after its Latin American exit. It does not create one uniform regulatory environment. Pennsylvania, Washington, and Quebec have different utilities, permitting systems, and energy markets.

The company’s financing position provides a cushion against delays, but debt introduces obligations. The June notes mature in 2032 and carry interest. Conversion can also dilute existing shareholders under specified conditions.

Management’s claim that it can finance construction smoothly remains forward-looking. The evidence required is concrete: signed leases, committed project financing, construction notices, equipment installation, and ready-for-service dates.

This skeptical view does not require rejecting Keel’s power thesis. Grid-connected locations have real strategic value in an electricity-constrained market. The question is how much value Keel can retain after development costs, financing, and tenant negotiations.

The safest interpretation is conditional. Keel has assembled a credible platform for pursuing AI infrastructure. The platform becomes a functioning business only when customers sign, projects receive financing, and sites deliver usable capacity.

Three Signals Will Show Whether Keel Can Deliver

The next stage should be judged through leases, construction milestones, and financial conversion, in that order.

The first signal is a binding tenant agreement at Panther Creek, Sharon, or Moses Lake. Management says negotiations are active across all three sites. A signed lease would provide the strongest external validation of Keel’s repositioning.

The details will matter. Investors should look for the customer’s credit quality, contracted capacity, lease duration, service date, and financing obligations. An unnamed preliminary agreement would carry less weight than an enforceable contract with clear milestones.

A lease at Moses Lake would validate the smaller, more advanced location. A Panther Creek agreement would support the scale of Keel’s flagship campus. Sharon could test whether the company can combine existing operations with new substation capacity.

The second signal is physical delivery against the announced schedules. Keel’s earlier materials listed 2027 as the earliest ready-for-service year for Panther Creek, Sharon, and Moses Lake. Sherbrooke followed with an earliest date in 2028.

Ready for service means the customer can begin using contracted infrastructure after required testing and acceptance. It is more meaningful than a groundbreaking ceremony or equipment delivery.

Progress should include final permits, completed fiber contracts, utility milestones, construction activity, and installation of critical systems. Repeated schedule changes would weaken confidence, even if long-term demand remains strong.

The Vertiv equipment at Moses Lake offers an early test. Investors should watch whether those modules move from delivery into installed, commissioned infrastructure. The difference separates procurement progress from operating capacity.

The third signal is financial conversion from shrinking mining revenue toward contracted infrastructure economics. Keel’s legacy revenue fell sharply during the second quarter. New AI revenue needs to replace that decline before liquidity becomes the entire story.

Useful indicators include project financing commitments, construction spending, remaining liquidity, customer deposits, and the timing of recurring lease revenue. Adjusted metrics should be compared with standard accounting results.

Keel’s $819 million liquidity balance currently supports negotiations and development. That advantage weakens if permitting or leasing takes much longer than planned. It strengthens if contracts unlock project-level financing before major construction spending accelerates.

The company’s remaining Bitcoin position also deserves attention. Selling more Bitcoin can fund development, but it reduces a liquid reserve whose value changes with the cryptocurrency market. Keeping it preserves exposure to the business Keel is leaving.

Google News will continue surfacing each appointment, permit, financing, and equipment delivery. Those updates provide useful pieces, but none should replace the three-part test.

First, a credible customer must sign. Second, Keel must build and commission the promised capacity. Third, that capacity must generate returns after construction and financing costs.

Keel has already crossed a meaningful threshold by decommissioning U.S. mining operations. It has chosen the AI infrastructure route and committed capital to it. Reversing course would now carry operational and reputational costs.

The next question belongs to customers as much as investors. Will an AI operator trust Keel with a high-density workload and a fixed delivery schedule? That decision would test the company more rigorously than another corporate announcement.

Readers following the transition should save the August baseline and compare every later update against it. Track each site independently, record changes in capacity or timing, and distinguish negotiations from executed contracts.

The Google News narrative says a Bitcoin miner became an AI infrastructure company. The evidence supports only the first half today. Keel has left much of mining behind, but leases and delivered capacity must complete the second half.